Parallelization of discrete cosine transform and VP8 algorithms on GPU for lossy image compression

V diplomskem delu preučimo uporabo grafično procesne enote (GPE) za namen kompresije slik. Najprej teoretično zasnujemo algoritma VP8 in diskretno kosinusno transformacijo (DCT), nato predstavimo njuno osnovno implementacijo. Algoritem DCT še dodatno optimiziramo in oba algoritma paraleliziramo z uporabo GPE. Na koncu primerjamo hitrost obeh implementacij in ugotovimo kdaj je za stiskanje slik smiselno uporabiti GPE.In this thesis we study the use of graphics processing units (GPU) for image compression. We begin by theoretically defining algorithms VP8 and discrete cosine transform (DCT), then we present a basic implementation of the two algorithms. Additionaly, we optimize the DCT implementation and parallelize both algorithms to be used on a GPU. Finally, we compare the performance of both implementations and determine when using GPUs for image compression is suitable