MRES.B.02.01 Selected Topics in image Processing and Computer Vision (MRES.B.02.01)

The sparse representation (SR) is a neurological inspired computer vision technique in which only a number of few vision elements - receptive fields are activated. Thus, sparse coding occurs when an image (object) under examination is encoded by the strong activation of a relatively small set of neurons. For each image to be encoded, there is a  completely different subset of all available neurons. 

In computer vision,the k-SVD is a dictionary (i.e. codebook) learning (DL) algorithm which creates one dictionary D, employed in sparse representation coding. k-SVD, is a merging of the singular value decomposition (SVD) and a generalization of the k-means clustering method. The k-SVD algorithm has been applied to numerous applications such as image denoising and recognition, audio processing, biology, and document analysis.

The k-SVD is always accompanied by its Orthogonal Matching Pursuit (OMP) algorithm counterpart. The OMP is actually a sparse representation algorithm, which finds the "best matching" projections of multidimensional data onto the span of an over-complete  dictionary D. 

Both k-SVD and OMP are considered to be examples of joint optimization allgorithms.