Capsule endoscopy is a method for recording images of the digestive tract. A patient swallows a capsule containing a tiny camera, which captures images that are then transmitted wirelessly to an external receiver for examination by a physician. Due to limited computational capabilities in the capsule and bandwidth constraints derived from dimensions of the capsule, low-complexity and efficient compression of the images is required before transmission.
The goal of this project is to use sparse representations in the compression scheme and evaluate the resulting compression ratios relative to other compression methods. We begin with learning a sparse dictionary with the K-SVD algorithm. Then we use the Orthogonal Matching Pursuit (OMP) sparse coding algorithm in order to express the endoscopic images as sparse vectors. Following, we apply the stages of quantization and entropy coding to obtain the final stream that is transmitted from the capsule. As opposed to other compression methods like JPEG, in sparse coding compression we are also required to encode the indices of dictionary atoms that were used in the sparse coding. This encoding needs to be lossless, so compression ratio problems can arise from this stage of our tested sparse coding based compression scheme. The results surpass the JPEG algorithm up to 40 dB, but at the required rate of 42.5 dB, they are inferior to it.