Emprical Study and Development of Image Enhancement Techniques

Abstract

Digital Image quality measurement is considered as the most considerable feature in image processing systems and especially for efficient transmission. While performing image compression, blocking effect is produced due to co-efficient quantification that in turn results in poor quality of decompressed image. In addition, the most common shortcomings of these methods are the lack of non-linear de-blocking of artifacts over multiple scales. Hence, quality enhancement forms an essential and prerequisite process in image processing for several applications, namely, medical images, traffic surveillance and so on. newlineRecently, many research works have been designed for quality evaluation of digital images. The performances of existing techniques were not sufficient due to blocking and de-blocking artifacts. In order to overcome above such issues, three different methods such as PN-DB framework, NDC-DQE technique and MFNBGO technique are developed with various numbers of images from digital image database. newlineThe proposed PN-DB framework, NDC-DQE technique and MFNBGO technique is performed by using Manuscripts and Archives Digital Images Database (MADID) to validate the simulation process. The MADID database includes with digital images in the form of photographs, posters, drawings, text documents and other images. It is the research collections of Manuscripts and Archives, Yale University Library. The MADID comprises manuscript group number followed by collection name which includes a title, image number, original material, copyright holder, description, manuscript group name, manuscript box number, folder number, folder name, file name and so on. The digital images with various sizes from MADID, Yale University Library are taken for conducting the simulation work of proposed methods. From MADID, digital images in the range between 10 to 100 is considered with sizes ranges between 373 MB to 985 MB are taken for experimental purpose. newline