Useful features for classifying color images -
Useful features for classifying color images -
i asked question on signal processing so. reposting here didn't response there.
i have 2 sets of images. in 1 set, background , foreground differs in color (either background dark , foreground lite or vice versa). useful features distinguishing between these types of images?
i started histograms of colors (i concatenating r/g/b histograms 8 bins 24 dimensional feature vector), can improved? using other color spaces?
edit
i have added image example. greenish rectangle shows part colors not vary much. magenta rectangle shows part do. separate these 2 regions. doing calculating histograms of color channels on these regions , concatenating them.
if looking regions colours vary, suggest averaging colours , subtracting each pixel avergae of pixels in vicinity. image homogeneous, difference between pixel , average of local area tend 0 - i.e. black.
here have done gaussian blur 5 pixel radius , differenced relative original image, i.e. subtracted blurred image original.
i can accomplish similar result using imagemagick this, in greyscale:
convert vbhc1.png -colorspace grayness \( +clone -blur 0x6 \) -compose difference -composite out.jpg
or maybe border detector, such canny tell if in desired area or not. made imagemagick using
convert vbhc1.png -canny 0x1+10%+20% out.jpg
image image-processing machine-learning
Comments
Post a Comment