Fen ecosystem carbon gas dynamics in changing hydrological conditions

Northern peatlands are thought to store one third of all soil carbon (C). Besides the C sink function, peatlands are one of the largest natural sources of methane (CH4) to the atmosphere. Climate change may affect the C gas dynamics as well as the labile C pool. Because the peatland C sequestration and CH4 emissions are governed by high water levels, changes in hydrology are seen as the driving factor in peatland ecosystem change. This study aimed to quantify the carbon dioxide (CO2) and CH4 dynamics of a fen ecosystem at different spatial scales: plant community components scale, plant community scale and ecosystem scale, under hydrologically normal and water level drawdown conditions. C gas exchange was measured in two fens in southern Finland applying static chamber and eddy covariance techniques. During hydrologically normal conditions, the ecosystem was a CO2 sink and CH4 source to the atmosphere. Sphagnum mosses and sedges were the most important contributors to the community photosynthesis. The presence of sedges had a major positive impact on CH4 emissions while dwarf shrubs had a slightly attenuating impact. C fluxes varied considerably between the plant communities. Therefore, their proportions determined the ecosystem scale fluxes. An experimental water level drawdown markedly reduced the photosynthesis and respiration of sedges and Sphagnum mosses and benefited shrubs. Consequently, changes were smaller at the ecosystem scale than at the plant group scale. The decrease in photosynthesis and the increase in respiration, mostly peat respiration, made the fen a smaller CO2 sink. CH4 fluxes were significantly lowered, close to zero. The impact of natural droughts was similar to, although more modest than, the impact of the experimental water level drawdown. The results are applicable to the short term impacts of the water level drawdown and to climatic conditions in which droughts become more frequent.

The role of pollen in the changing environmental conditions of Scots pine

Rapid change in climate is challenge for the adaptation of forest trees in the future. In wind pollinated tree species pollen mediated long distance gene flow may provide alleles that are (pre)adapted to a future climate. In order to examine the long distance pollen flow in Scots pine (Pinus sylvestris L.), we measured the amount and viability of airborne pollen and flowering phenology in central, northern, and northernmost Finland during four years. Viable airborne pollen grains were detected during female flowering and before local pollen shedding in all study sites. The situation when there was nonlocal pollen in the air lasted from one to four days depending on the year and study site. The amount of nonlocal airborne pollen varied also between years and study sites, the total amount of nonlocal viable pollen in the air was 2.3% from all detected viable pollen grains. The effect of pollen origin on seeds siring ability was studied with artificial pollination experiments. Pollen genotypes originating from southern Finland sired 76% and 48 % of the analysed seeds in competition studies where both pollen origin were introduced simultaneously into the female strobili. We examined the importance of arrival order of pollen grains in to the strobili in a study where pollen genotypes of different origin were introduced in two hours interval. Northern genotypes sired 76% of the analysed seeds when it was injected first, but in the "southern first" experiment both pollen types sired equal amount of seeds. The first pollen grain in the pollen chamber do not always fertilizes the ovum, instead there likely is more complex way of competition between pollen grains. To examine chemically mediated pollen-pollen interactions we conducted in vitro germination experiment where different pollen genotypes had chemical but not physical contact. Both positive and negative effects of interactions were found. We found highly negative effects in germinability of northern pollen grains when they were germinating with southern pollen, and increase in the germinability of southern pollen. There were no variation in the size of the dry pollen grains between pollen origins, and minor variation between different genotypes. After hydration and germination northern pollen grains were larger than southern pollen. Pollen genotypes having high hydration rates had low germinability and tube growth rate, however, germinated pollen grains were larger in size than nongerminated. This supports the suggestion that the early germination and growth of pollen tube is dependent on pollen storage materialsand less dependent on water intake and hydration. Long distance pollen movements and good competition ability of southern pollen makes gene flow possible, although rising temperature and timing of pollen movements may affect pollen competition and the amount of gene flow.

Soil hydrological properties and conditions, site preparation, and the long-term performance of planted Scots pine (Pinus sylvestris L.) on upland forest sites in Finnish Lapland

Scots pine (Pinus sylvestris L.) and Norway spruce (Picea abies (L.) Karst.) forests dominate in Finnish Lapland. The need to study the effect of both soil factors and site preparation on the performance of planted Scots pine has increased due to the problems encountered in reforestation, especially on mesic and moist, formerly spruce-dominated sites. The present thesis examines soil hydrological properties and conditions, and effect of site preparation on them on 10 pine- and 10 spruce-dominated upland forest sites. Finally, the effects of both the site preparation and reforestation methods, and soil hydrology on the long-term performance of planted Scots pine are summarized. The results showed that pine and spruce sites differ significantly in their soil physical properties. Under field capacity or wetter soil moisture conditions, planted pines presumably suffer from excessive soil water and poor soil aeration on most of the originally spruce sites, but not on the pine sites. The results also suggested that site preparation affects the soil-water regime and thus prerequisites for forest growth over two decades after site preparation. High variation in the survival and mean height of planted pine was found. The study suggested that on spruce sites, pine survival is the lowest on sites that dry out slowly after rainfall events, and that height growth is the fastest on soils that reach favourable aeration conditions for root growth soon after saturation, and/or where the average air-filled porosity near field capacity is large enough for good root growth. Survival, but not mean height can be enhanced by employing intensive site preparation methods on spruce sites. On coarser-textured pine sites, site preparation methods don t affect survival, but methods affecting soil fertility, such as prescribed burning and ploughing, seem to enhance the height growth of planted Scots pines over several decades. The use of soil water content in situ as the sole criterion for sites suitable for pine reforestation was tested and found to be a relatively uncertain parameter. The thesis identified new potential soil variables, which should be tested using other data in the future.

Expression Analysis of Host-Induced Genes and Proteins of Potato Pathogen Pectobacterium atrosepticum

Pectobacterium atrosepticum on Gram-negatiivinen bakteeri, joka aiheuttaa perunan tyvi- ja märkämätää. P. atrosepticum bakteerin optimilämpötila on melko alhainen ja se on yleinen lauhkeilla alueilla. Tyvimätä leviää pääasiassa siemenperunan välityksellä ja siksi se on ongelma erityisesti siemenperunan tuotannossa. P. atrosepticum kannan SCRI1043 genomi on julkaistu ja sitä tutkitaan malliorganismina märkä- ja tyvimädän taudinaiheuttamisen ymmärtämiseksi. Tämä opportunistinen taudinaiheuttaja voi elää isäntäkasvissa kuukausia piilevänä, aiheuttamatta näkyviä oireita. Suotuisissa olosuhteissa bakteerit alkavat jakautua ja tuottaa kasvin kudoksia hajottavia entsyymejä. Mädäntyvä kasvimassa tarjoaa ravinteita bakteerien kasvuun ja mahdollistaa isäntäkasvin asuttamisen. Soluseiniä hajottavien entsyymien merkitys taudinaiheuttamisessa on hyvin tunnettu, mutta oireettomasta jaksosta ja taudin alkuvaiheista tiedätään vain vähän. Bakteerin genomi sisältää monia toksiineja, adhesiineja, hemolysiineja ja muita proteiineja, joilla saattaa olla merkitys taudinaiheuttamisessa. Tässä työssä käytettiin proteomiikkaa ja mikrosiruanalysiä P. atrosepticum bakteerin erittyvien proteiinien ja geeniekspression tutkimiseen. Proteiinit, jotka eritetään ulos bakteerista, toimivat todennäköisesti taudinaiheuttamisessa, koska ne ovat suorassa kontaktissa isäntäkasvin kanssa. Analyysit suoritettiin olosuhteissa, jotka muistuttavat kasvin soluvälitilaa: matala pH, vähän ravinteita ja matala lämpötila. Isäntäkasvin läsnäolon vaikutusta proteiinien tuottoon ja geeniekspressioon tutkittiin lisäämällä perunauutetta kasvatusalustaan. Tutkimuksessa tunnistettiin P. atrosepticum bakteerin monia jo tunnettuja ja mahdollisesti taudinaiheuttamiseen liittyviä proteiineja. Perunauute lisäsi hiljattain tunnistetun, proteiinien eritysreittiä (tyyppi VI sekreetio, T6SS) koodaavien geenien ilmentymistä. Lisäksi bakteerin havaittiin erittävän useita T6SS:n liittyviä proteiineja kasvualustaan, johon oli lisätty perunauutetta. T6SS:n merkitys bakteereille on vielä epäselvä ja sen vaikutuksesta taudinaiheuttamiseen on julkaistu ristiriitaisia tuloksia. Märkä- ja tyvimädän ymmärtäminen molekulaarisella tasolla luo pohjan tautien kontrollointiin tähtäävään soveltavaan tutkimukseen. Tämä tutkimus lisää tietoa kasvi-patogeeni- interaktiosta ja sitä voidaan tulevaisuudessa käyttää hyväksi esimerkiksi diagnostiikassa, resistenttien perunalajikkeiden jalostuksessa tai viljely- ja varastointiolosuhteiden parantamisessa.

Nitrogen dynamics and nitrogen use efficiency of spring cereals under Finnish growing conditions

Nitrogen (N) is one of the main inputs in cereal cultivation and as more than half of the arable land in Finland is used for cereal production, N has contributed substantially to agricultural pollution through fertilizer leaching and runoff. Based on this global phenomenon, the European Community has launched several directives to reduce agricultural emissions to the environment. Trough such measures, and by using economic incentives, it is expected that northern European agricultural practices will, in the future, include reduced N fertilizer application rates. Reduced use of N fertilizer is likely to decrease both production costs and pollution, but could also result in reduced yields and quality if crops experience temporary N deficiency. Therefore, more efficient N use in cereal production, to minimize pollution risks and maximize farmer income, represents a current challenge for agronomic research in the northern growing areas. The main objective of this study was to determine the differences in nitrogen use efficiency (NUE) among spring cereals grown in Finland. Additional aims were to characterize the multiple roles of NUE by analysing the extent of variation in NUE and its component traits among different cultivars, and to understand how other physiological traits, especially radiation use efficiency (RUE) and light interception, affect and interact with the main components of NUE and contribute to differences among cultivars. This study included cultivars of barley (Hordeum vulgare L.), oat (Avena sativa L.) and wheat (Triticum aestivum L.). Field experiments were conducted between 2001 and 2004 at Jokioinen, in Finland. To determine differences in NUE among cultivars and gauge the achievements of plant breeding in NUE, 17-18 cultivars of each of the three cereal species released between 1909 and 2002 were studied. Responses to nitrogen of landraces, old cultivars and modern cultivars of each cereal species were evaluated under two N regimes (0 and 90 kg N ha-1). Results of the study revealed that modern wheat, oat and barley cultivars had similar NUE values under Finnish growing conditions and only results from a wider range of cultivars indicated that wheat cultivars could have lower NUE than the other species. There was a clear relationship between nitrogen uptake efficiency (UPE) and NUE in all species whereas nitrogen utilization efficiency (UTE) had a strong positive relationship with NUE only for oat. UTE was clearly lower in wheat than in other species. Other traits related to N translocation indicated that wheat also had a lower harvest index, nitrogen harvest index and nitrogen remobilisation efficiency and therefore its N translocation efficiency was confirmed to be very low. On the basis of these results there appears to be potential and also a need for improvement in NUE. These results may help understand the underlying physiological differences in NUE and could help to identify alternative production options, such as the different roles that species can play in crop rotations designed to meet the demands of modern agricultural practices.

Winter feeding strategies for suckler cows in cold climatic conditions

In Finland, suckler cow production is carried out in circumstances characterized by a long winter period and a short grazing period. The traditional winter housing system for suckler cows has been insulated or uninsulated buildings, but there is a demand for developing less expensive housing systems. In addition, more information is needed on new winter feeding strategies, carried out in inexpensive winter facilities with conventional (hay, grass silage, straw) or alternative (treated straw, industrial by-product, whole-crop silage) feeds. The new feeding techniques should not have any detrimental effects on animal welfare in order to be acceptable to both farmers and consumers. Furthermore, no official feeding recommendations for suckler cows are available in Finland and, thus, recommendations for dairy cows have been used. However, this may lead to over- or underfeeding of suckler cows and, finally, to decreased economic output. In Experiment I, second-calf beef-dairy suckler cows were used to compare the effects of diets based on hay (H) or urea-treated straw (US) at two feeding levels (Moderate; M vs. Low; L) on the performance of cows and calves. Live weight (LW) gain during the indoor feeding was lower for cows on level L than on level M. Cows on diet US lost more LW indoors than those on diet H. The cows replenished the LW losses on good pasture. Calf LW gain and cow milk production were unaffected by the treatments. Conception rate was unaffected by the treatments but was only 69%. Urea-treated straw proved to be a suitable winter feed for spring-calving suckler cows. Experiment II studied the effects of feeding accuracy on the performance of first- and second-calf beef-dairy cows and calves. In II-1, the day-to-day variation in the roughage offered ranged up to ± 40%. In II-2, the same variation was used in two-week periods. Variation of the roughages offered had minor effects on cow performance. Reproduction was unaffected by the feeding accuracy. Accurate feeding is not necessary for young beef-dairy crosses, if the total amount of energy offered over a period of a few weeks fulfills the energy requirements. Effects of feeding strategies with alternative feeds on the performance of mature beef-dairy and beef cows and calves were evaluated in Experiment III. Two studies consisted of two feeding strategies (Step-up vs. Flat-rate) and two diets (Control vs. Alternative). There were no differences between treatments in the cow LW, body condition score (BCS), calf pre-weaning LW gain and cow reproduction. A flat-rate strategy can be practised in the nutrition of mature suckler cows. Oat hull based flour-mill by product can partly replace grass silage and straw in the winter diet. Whole-crop barley silage can be offered as a sole feed to suckler cows. Experiment IV evaluated during the winter feeding period the effects of replacing grass silage with whole-crop barley or oat silage on mature beef cow and calf performance. Both whole-crop silages were suitable winter feeds for suckler cows in cold outdoor winter conditions. Experiment V aimed at assessing the effects of daily feeding vs. feeding every third day on the performance of mature beef cows and calves. No differences between the treatments were observed in cow LW, BCS, milk production and calf LW. The serum concentrations of urea and long-chain fatty acids were increased on the third day after feeding in the cows fed every third day. Despite of that the feeding every third day is an acceptable feeding strategy for mature suckler cows. Experiment VI studied the effects of feeding levels and long-term cold climatic conditions on mature beef cows and calves. The cows were overwintered in outdoor facilities or in an uninsulated indoor facility. Whole-crop barley silage was offered either ad libitum or restricted. All the facilities offered adequate shelter for the cows. The restricted offering of whole-crop barley silage provided enough energy for the cows. The Finnish energy recommendations for dairy cows were too high for mature beef breed suckler cows in good body condition at housing, even in cold conditions. Therefore, there is need to determine feeding recommendations for suckler cows in Finland. The results showed that the required amount of energy can be offered to the cows using conventional or alternative feeds provided at a lower feeding level, with an inaccurate feeding, flat-rate feeding or feeding every third day strategy. The cows must have an opportunity to replenish the LW and BCS losses at pasture before the next winter. Production in cold conditions can be practised in inexpensive facilities when shelter against rain and wind, a dry resting place, adequate amounts of feed suitable for cold conditions and water are provided for the animals as was done in the present study.

Digestion capasity, nutrient digestibilities and physico-chemical conditions in the intestine influenced by the age of growing turkeys

Six experiments have been conducted to examine digestibility and feeding value of domestic Finnish fibre-rich cereals (barley and oats as compared to maize and wheat) and protein sources (rapeseed meal and cake, peas, faba beans, lupin seeds) for growing turkeys and to investigate effects of age of the birds (from 3 to 12 weeks of age) on digestion process and estimated nutrient digestibility and energy values. Besides, an objective of the study was to test applications of digestibility research methodology for turkeys. Total tract digestibility and apparent metabolizable energy (AME) was assayed in experimental cages using excreta collection, and a slaughter method was applied to sample small intestinal digesta for determination of apparent ileal crude protein digestibility (AICPD), jejuno-duodenal digesta viscosity and caecal volatile fatty acid (VFA) concentration. Digesta viscosity decreased and caecal VFA production increased with age of growing turkeys. Digesta retention times in the small intestine were generally longer in the older birds than in the younger ones. Crude fat digestibility and AME increased with age of growing turkeys, especially with viscous diets. AICPD seemed to decrease with age in most cases. Supplementation with β-gucanase-xylanase decreased viscosity, improved crude fat digestibility and metabolizable energy value and increased VFA production especially in barley-fed turkeys and especially in the young birds. Poor protein digestibility and low energy value of rapeseed meal and rapeseed cake decreased their feeding value for turkeys. In addition, a typical goitrogenic effect of rapeseed feeding was detected. Use of legume seeds as feed for growing turkeys is limited mostly by the low energy value in lupin seeds and the low ileal protein and amino acid digestibility in faba beans. Digestibility of fibre-rich protein sources was not improved with age of the turkeys. Euthanizing the turkeys for AICPD determination by carbon dioxide and bleeding led to lower digestibility values than mechanical stunning and cervical dislocation, suggesting inferiority of carbon dioxide stunning in experimental use. Comparison of AICPD and AME results obtained using different markers showed that considerable differences may occur, especially on total tract level, when acid-insoluble ash gave considerably lower AME values than titanium dioxide and chromic oxide.

Chemical characteristics and behaviour of sediment phosphorus in the northeastern Baltic Sea.

Chemical characteristics and behaviour of sediment phosphorus in the northeastern Baltic Sea Eutrophication is a severe environmental problem in the Baltic Sea, especially in the Gulf of Finland and the Archipelago Sea, and it is enhanced by the release of phosphorus (P) from bottom sediments. The release of P from sediment reserves largely depends on the occurrence of P in different chemical forms and on the prevailing conditions, especially on the presence of oxygen. This study examines the chemical character and the vertical distribution of sediment P in two shallow estuaries, in shallow coastal sediments overlain by oxic near-bottom water, and in poorly oxygenated open sea sediments in the northeastern Baltic Sea. The objective was to evaluate how much of the sediment P is buried and removed from the nutrient cycle, and how much of it is in forms that can be released from the sediment to the overlaying water over time. Relationships between the distribution of the different P forms and the chemical and physical properties of the sediment, sediment pore water, and near-bottom water were determined in order to examine the behaviour of P at the sediment-water interface. The results show that the chemical character of sediment P varied in the different areas. Generally, in the outer estuaries and in the organic-rich coastal areas in the eastern Gulf of Finland, the sediments were higher in P than the sediments in the poorly oxygenated open sea areas in the central and western Gulf. The estuary sediments that received erosion-transported material were characterised by P bound to hydrated oxides of iron and aluminum. Iron-bound P is sensitive to changes in redox-conditions, but part of it was buried in the estuaries, possibly because of high sedimentation rates and incomplete reduction of iron. The open sea sediments in the central and western Gulf of Finland were dominated by apatite-P, which was also abundant in the areas strongly affected by sediment transportation. The burial of sediment P was most effective in the areas rich in apatite-P, which is a relatively stable form of P in sediment. In the eastern Gulf of Finland, organic P forms predominated in the organic-rich sediments. A part of these P forms will be buried, while part will be degraded in the long term, releasing soluble P to the pore water. In the poorly oxygenated areas, iron compounds at the sediment surface are not able to retain P released during mineralisation of organic matter or reduction-induced dissolution of iron-compounds in deep sediment layers. However, in the shallow coastal areas overlain by oxic near-bottom water, the organic-rich surface sediment can also become temporarily reduced and release P from the sediment to the overlaying water. The considerable variation in the chemical composition of sediment P reserves in the northeastern Baltic Sea proved that it is an important factor and should be taken into account when evaluating the release of sediment P and the role of P reserves in bottom sediments in eutrophication.

Capillary electrochromatography: a versatile instrumental technique for nanodomain interaction studies

This doctoral thesis describes the development of a miniaturized capillary electrochromatography (CEC) technique suitable for the study of interactions between various nanodomains of biological importance. The particular focus of the study was low-density lipoprotein (LDL) particles and their interaction with components of the extracellular matrix (ECM). LDL transports cholesterol to the tissues through the blood circulation, but when the LDL level becomes too high the particles begin to permeate and accumulate in the arteries. Through binding sites on apolipoprotein B-100 (apoB-100), LDL interacts with components of the ECM, such as proteoglycans (PGs) and collagen, in what is considered the key mechanism in the retention of lipoproteins and onset of atherosclerosis. Hydrolytic enzymes and oxidizing agents in the ECM may later successively degrade the LDL surface. Metabolic diseases such as diabetes may provoke damage of the ECM structure through the non-enzymatic reaction of glucose with collagen. In this work, fused silica capillaries of 50 micrometer i.d. were successfully coated with LDL and collagen, and steroids and apoB-100 peptide fragments were introduced as model compounds for interaction studies. The LDL coating was modified with copper sulphate or hydrolytic enzymes, and the interactions of steroids with the native and oxidized lipoproteins were studied. Lipids were also removed from the LDL particle coating leaving behind an apoB-100 surface for further studies. The development of collagen and collagen decorin coatings was helpful in the elucidation of the interactions of apoB-100 peptide fragments with the primary ECM component, collagen. Furthermore, the collagen I coating provided a good platform for glycation studies and for clarification of LDL interactions with native and modified collagen. All methods developed are inexpensive, requiring just small amounts of biomaterial. Moreover, the experimental conditions in CEC are easily modified, and the analyses can be carried out in a reasonable time frame. Other techniques were employed to support and complement the CEC studies. Scanning electron microscopy and atomic force microscopy provided crucial visual information about the native and modified coatings. Asymmetrical flow field-flow fractionation enabled size measurements of the modified lipoproteins. Finally, the CEC results were exploited to develop new sensor chips for a continuous flow quartz crystal microbalance technique, which provided complementary information about LDL ECM interactions. This thesis demonstrates the potential of CEC as a valuable and flexible technique for surface interaction studies. Further, CEC can serve as a novel microreactor for the in situ modification of LDL and collagen coatings. The coatings developed in this study provide useful platforms for a diversity of future investigations on biological nanodomains.

TlBr raw material purification, crystal growth, annealing, detector fabrication and characterisation for gamma-ray detector applications

The research reported in this thesis dealt with single crystals of thallium bromide grown for gamma-ray detector applications. The crystals were used to fabricate room temperature gamma-ray detectors. Routinely produced TlBr detectors often are poor quality. Therefore, this study concentrated on developing the manufacturing processes for TlBr detectors and methods of characterisation that can be used for optimisation of TlBr purity and crystal quality. The processes under concern were TlBr raw material purification, crystal growth, annealing and detector fabrication. The study focused on single crystals of TlBr grown from material purified by a hydrothermal recrystallisation method. In addition, hydrothermal conditions for synthesis, recrystallisation, crystal growth and annealing of TlBr crystals were examined. The final manufacturing process presented in this thesis deals with TlBr material purified by the Bridgman method. Then, material is hydrothermally recrystallised in pure water. A travelling molten zone (TMZ) method is used for additional purification of the recrystallised product and then for the final crystal growth. Subsequent processing is similar to that described in the literature. In this thesis, literature on improving quality of TlBr material/crystal and detector performance is reviewed. Aging aspects as well as the influence of different factors (temperature, time, electrode material and so on) on detector stability are considered and examined. The results of the process development are summarised and discussed. This thesis shows the considerable improvement in the charge carrier properties of a detector due to additional purification by hydrothermal recrystallisation. As an example, a thick (4 mm) TlBr detector produced by the process was fabricated and found to operate successfully in gamma-ray detection, confirming the validity of the proposed purification and technological steps. However, for the complete improvement of detector performance, further developments in crystal growth are required. The detector manufacturing process was optimized by characterisation of material and crystals using methods such as X-ray diffraction (XRD), polarisation microscopy, high-resolution inductively coupled plasma mass (HR-ICPM), Fourier transform infrared (FTIR), ultraviolet and visual (UV-Vis) spectroscopy, field emission scanning electron microscope (FESEM) and energy-dispersive X-ray spectroscopy (EDS), current-voltage (I-V) and capacity voltage (CV) characterisation, and photoconductivity, as well direct detector examination.

A Probabilistic Approach to the Primary Visual Cortex

What can the statistical structure of natural images teach us about the human brain? Even though the visual cortex is one of the most studied parts of the brain, surprisingly little is known about how exactly images are processed to leave us with a coherent percept of the world around us, so we can recognize a friend or drive on a crowded street without any effort. By constructing probabilistic models of natural images, the goal of this thesis is to understand the structure of the stimulus that is the raison d etre for the visual system. Following the hypothesis that the optimal processing has to be matched to the structure of that stimulus, we attempt to derive computational principles, features that the visual system should compute, and properties that cells in the visual system should have. Starting from machine learning techniques such as principal component analysis and independent component analysis we construct a variety of sta- tistical models to discover structure in natural images that can be linked to receptive field properties of neurons in primary visual cortex such as simple and complex cells. We show that by representing images with phase invariant, complex cell-like units, a better statistical description of the vi- sual environment is obtained than with linear simple cell units, and that complex cell pooling can be learned by estimating both layers of a two-layer model of natural images. We investigate how a simplified model of the processing in the retina, where adaptation and contrast normalization take place, is connected to the nat- ural stimulus statistics. Analyzing the effect that retinal gain control has on later cortical processing, we propose a novel method to perform gain control in a data-driven way. Finally we show how models like those pre- sented here can be extended to capture whole visual scenes rather than just small image patches. By using a Markov random field approach we can model images of arbitrary size, while still being able to estimate the model parameters from the data.

Learning Nonlinear Visual Processing from Natural Images

The paradigm of computational vision hypothesizes that any visual function -- such as the recognition of your grandparent -- can be replicated by computational processing of the visual input. What are these computations that the brain performs? What should or could they be? Working on the latter question, this dissertation takes the statistical approach, where the suitable computations are attempted to be learned from the natural visual data itself. In particular, we empirically study the computational processing that emerges from the statistical properties of the visual world and the constraints and objectives specified for the learning process. This thesis consists of an introduction and 7 peer-reviewed publications, where the purpose of the introduction is to illustrate the area of study to a reader who is not familiar with computational vision research. In the scope of the introduction, we will briefly overview the primary challenges to visual processing, as well as recall some of the current opinions on visual processing in the early visual systems of animals. Next, we describe the methodology we have used in our research, and discuss the presented results. We have included some additional remarks, speculations and conclusions to this discussion that were not featured in the original publications. We present the following results in the publications of this thesis. First, we empirically demonstrate that luminance and contrast are strongly dependent in natural images, contradicting previous theories suggesting that luminance and contrast were processed separately in natural systems due to their independence in the visual data. Second, we show that simple cell -like receptive fields of the primary visual cortex can be learned in the nonlinear contrast domain by maximization of independence. Further, we provide first-time reports of the emergence of conjunctive (corner-detecting) and subtractive (opponent orientation) processing due to nonlinear projection pursuit with simple objective functions related to sparseness and response energy optimization. Then, we show that attempting to extract independent components of nonlinear histogram statistics of a biologically plausible representation leads to projection directions that appear to differentiate between visual contexts. Such processing might be applicable for priming, \ie the selection and tuning of later visual processing. We continue by showing that a different kind of thresholded low-frequency priming can be learned and used to make object detection faster with little loss in accuracy. Finally, we show that in a computational object detection setting, nonlinearly gain-controlled visual features of medium complexity can be acquired sequentially as images are encountered and discarded. We present two online algorithms to perform this feature selection, and propose the idea that for artificial systems, some processing mechanisms could be selectable from the environment without optimizing the mechanisms themselves. In summary, this thesis explores learning visual processing on several levels. The learning can be understood as interplay of input data, model structures, learning objectives, and estimation algorithms. The presented work adds to the growing body of evidence showing that statistical methods can be used to acquire intuitively meaningful visual processing mechanisms. The work also presents some predictions and ideas regarding biological visual processing.