Reconhecimento de faces sob diferentes condições de iluminação utilizando PCA e a transformada census

Os seres humanos identificam naturalmente outros seres humanos utilizando suas características físicas, fisiológicas ou comportamentais. Dentre essas características, destacam-se os traços faciais. O avanço tecnológico na área de Biometria tem promovido o desenvolvimento de inúmeras técnicas para o reconhecimento automático de faces por meio de computadores, entretanto, existem ainda vários fatores que dificultam esta aplicação, como por exemplo, a variação das condições de iluminação. O objetivo deste artigo é analisar os efeitos da aplicação de um filtro de processamento de imagens, denominado Transformada Census, em uma base de dados com imagens da face em diferentes condições de iluminação. Assim, experimentos foram realizados utilizando a técnica PCA com imagens da base de dados AR antes e depois da aplicação da Transformada Census. Os resultados desses experimentos mostraram que a aplicação da Transformada Census melhorou o resultado do reconhecimento das faces, reduzindo a taxa de erro.

Genomic gains in prostatic carcinoma and proliferative inflammatory atrophy in dogs

The dog can spontaneously develop prostate cancer and consequently can be used as an experimental model for prostatic diseases associated with aging, including benign prostate hyperplasia (BPH) and prostate carcinoma (PCa). DNA copy number variations (CNVs) have been used to identify genes associated with cancer development and progression. DNA microarray based comparative genomic hybridization (aCGH) is a technique that allows to identify copy number of thousands of genes throughout the genome. aCGH was used to identify genomic regions with significantly different DNA copy number in three benign prostatic hyperplasia (BPH), four proliferative inflammatory atrophy (PIA), and 14 canine prostate carcinoma (PCa). Five histologically normal prostate tissue were used as reference. Genomic DNA was extracted from formalin fixed and paraffin embedded samples and CNVs data was evaluated in Canine Genome CGH Microarray 4x44K (G2519F, Design ID021193, Agilent). Data analysis was performed using Genomic Workbench Standard Edition 5.0.14 (Agilent). PCa showed higher number of altered genes related to canonical diseases process, cellular functions and molecular pathways as well as greater inter-relationship between genes, compared with PIA and BPH. In conclusion, PCa showed a more complex genotype, being losses the most frequent genomic changes. Some discrepancies between genomic alterations in human and canine carcinomas may indicate the different clinical behavior of these tumors in these two species. In addition, it was observed was an ascending pattern of genomic complexity in BPH, PIA and CA consistent with a model of multistep tumor progression.