Compressed Sensing and Its Applications in Video Processing

My research aims at the decoding of purely compressed-sensed videos sequences. In my major work of Ph.D. studies, I considered block-based CS video transmission systems. In such systems, the transmitter/encoder partitions each video frame into non-overlapping blocks, and performs nothing more than compressed sensing acquisition of each block independently, without the benefits of the familiar sophisticated forms of video encoding. Therefore, the burden of quality video sequence reconstruction falls solely on the receiver side. I developed block-adaptive Karhunen-Loève (KL) bases for sparse representation [9]-[11]. In a sliding-window decoding approach, the KL basis Ψmt,KLT for block X m t is estimated from the previously reconstructed frames F̂k, k = t−1, t−2, ..., t−n, where n is the number of reference frames. Experimental results demonstrate that the proposed decoder outperforms significantly the conventional fixed basis intra-frame [12] and inter-frame [13], as well as the K-SVD [14], decoders. Performance is improved as the number of reference frames (what we call “decoder order”) increases, with order values in the range of two to ten appearing as a good compromise between computational complexity and reconstruction quality.