Using Probabilistic Models for Data Compression

Our research objective is to improve the Huffman coding efficiency by adjusting data using a Poisson distribution, which avoids undefined entropies too. The scientific value added our paper consists in fact of minimizing average length code words, greater absence applying distribution. Coding an error-free compression method, designed remove redundancy, yielding smallest number symbols per source symbol, practice can be represented intensity image or output mapping operation. We shall use images from PASCAL Visual Object Classes (VOC) evaluate methods. In work we 10,102 randomly chosen images, such that half them are for training, while other testing. VOC sets display significant variability regarding object size, orientation, pose, illumination, position and occlusion. composed 20 classes, respectively: aeroplane, bicycle, bird, boat, bottle, bus, car, motorbike, train, sofa, table, chair, tv/monitor, potted plant, person, cat, cow, dog, horse sheep. descriptors different objects compared give measurement their similarity. Image similarity important concept many applications. This focused on measure computer science domain, more specifically information retrieval mining. approach uses 64 each belonging training test set, therefore 64. finite memory (Markov), where its depends previous outputs. When dealing with large volumes data, effective increase Information Retrieval speed based Neural Networks as artificial intelligent technique.