/saliency

A saliency detection algorithm based on brightness and contrast

Primary LanguagePython

saliency

A saliency detection algorithm based on brightness and contrast : Method is based on obtaining salient parts of the images by brightness and contrast values. The aim is to identify the pixels with high contrast, in a way with a high standard deviation, and label them as salient pixels. The logic depends on the following assumption: If ǀf(i,j)-Jǀ ≥ α C then f(i,j) is a salient pixel. Where, f(i,j) is pixel value J is brightness C is contrast alpha is a constant between [0.8-3] After converting the image to greyscale, the brightness and contrast values of the image are calculated. Brightness is defined by average pixel level for each band in the image. Contrast is defined by root mean square for each band in the image. After this process the salient pixels are transformed into white (255) and the rest to black (0), which is basic global thresholding. The alpha(α) parameter should be adjusted for better results.