Merged region based image retrieval

Efficient histogram for region based image retrieval in the discrete cosine transform domain

“…We have made a comparison between our proposed method (PM) and many benchmark existing RBIR methods like: regional convolution mapping feature with integrated category matching (RCMF+ICM) [17], adaptive region matching (ARM) [25], SIMPLIcity [21], integrated region matching (IRM) [25], MN-MIN [23], Shape-adaptive-DCT [30]. We also compare our approach with CBIR approaches like DCT+SVD [10] and DCT [9].…”

“…As discussed in [21], due to the uncontrollable nature of images, extracting objects from images automatically and precisely is still beyond the state of the art of new computer vision techniques [20], [35]. However, other systems tend to partition an object into multiple regions, none of which is representative of the semantic object [17]. Therefore, it is often difficult for users to determine which regions should be used for search.…”

“…Region-based search systems attempt to fill gaps in content-based search systems [15]- [17]. A regionbased search system applies segmentation [18]- [20] to images to decompose it into regions, which matches Int J Artif Intell ISSN: 2252-8938 ❒ 547 objects if the segmentation is perfect [21].…”

“…On the other hand, the significant region-based image retrieval (SRBIR) model proposed by [26], identifies a region of importance in an image using a mechanism based on visual attention and represents this region using the color descriptors and the descriptors based on the curvelet transform. Recently, the integrated category matching (ICM) was adopted by Meng et al [17] for similarity measurement combined with a centroid-based significance index method to match all merged regions in images according to their importance. For feature extraction in ICM method, a regional convolution mapping feature (RCMF) based on the convolutional neural networks (CNN) [27] was used.…”

Texture and colour region separation based image retrieval using probability annular histogram and weighted similarity matching scheme

“…Recently, Meng et al . [25] have used a regional convolution mapping feature scheme along with convolution neural network (CNN) based approach for feature extraction and image classification in CBIR applications. These CNN based schemes work in two stages, initially these schemes perform image classification task.…”