We have developed a number of image color related filters while
providing solutions for some computer vision problems and developed
some others just for fun. Here, we present some of the interesting
- Contrast Enhancement: as the name implies increasing the contrast of an image.
- Style Transfer: transfering the color profile of an image to another image.
- Decolorization with Contrast Preservation: converting a colored image to grayscale without losing its contrast information
The images to the right are contrast enhanced versions of the left
images. Contrast enancement is performed locally and it takes about
60ms/megapixel. It's highly parallelizable and could be made to run
Style transfer is the process of likening the photographic style of an
image to another one. It's highly popular nowadays in Computation
Photography literature. Our results below are acquired by implementing
a popular novel algorithm. The first row below contains
images where all the other images are likened to
in the corresponding rows below.
Decolorization with Contrast Preservation
Normal gray scale conversion is performed by weighted sum of the color
channels. This, however, results in a loss of contrast in the gray
converted images as it is not possible to distinguish similar but
permuted valued colors (i.e red (255,0,0) and blue (0,255,0) will look
very much alike when converted to gray scale). Contrast preserved
decolorization approaches solve this problem by applying certain
constraints on the grayscale conversion process to preserve as much
contrast as possible. Below, we show some results of our
implementation: Left image is the original colored image, middle one
is the standard gray-scale converted image and the right one is the
result of decolorization with contrast preservation.