The searching method based on the content by fusion of texture feature and qavelet moment feature

摘要

PURPOSE: A method for searching images based on contents is provided to store and search more effective information than text based search by automatically extracting a visual characteristic related to image contents such as a color and a texture, constructing a database, and comparing texture and similarity. CONSTITUTION: The first moment and the second moment of a BDIP(Block Difference of Inverse Probabilities) are extracted from blocks having a high BDIP and a low BDIP using an area map(104) which divides an image having a texture characteristic of a BDIP and a VLCC(Variation of Local Correlation Coefficient) into 4x4 and 8x8 size according to entropy(105,106). After calculating a VLCC by one pixel(102), the first moment and the second moment of the VLCC are extracted from a high entropy block and a low entropy block using the area map(104,107,108). A moment is extracted using a BDIP moment, a VLCC moment, and a wavelet conversion(103,109).