Original link: http: //www.51jrft.com/dzxx/dzkxyjs/552.html
In recent years, with the rapid development of multimedia technology and Internet technology, the rapid expansion of digital image resources, in order to effectively manage and retrieve these resources, image retrieval technologies have emerged and become one of the focus areas of the image. A prominent problem with existing image retrieval is huge semantic gap between low-level image features and semantic level. ROI detection is one of the mainstream technology to make the semantic gap. Visual attention model is based on the human visual system theory is the image most likely to draw attention to the area of the method significance of the size of the general image to represent, therefore, study the visual attention model of great significance to the region of interest detected image. HM000060
This article describes the current development of image retrieval technology and the problems faced; summarizes progress in basic research model of visual attention and the main implementation method, according to human visual system theory, the detection region of interest based on the model of visual attention Program. The main contents comprising: region detection method (1) An improved evolutionary programming. Evolutionary programming method is improved, increasing the threshold processing, region growing, etc., to obtain a candidate area realistic objectives; (2) to adjust the color by the similarity distance, brightness, weight characteristic direction of the weight is significantly characterized in visual tasks, after the area enhancement factor correction Bottom-Up visual attention model detection region of interest; (3) the improved projection display model, the visual saliency multi-scale transformed by wavelet calculated using evolutionary programming global saliency, with Centre-surrounded calculating local saliency fusion obtained; factor on visual salience corrected based on visual impact, and the candidate region combined to give the region of interest by the salient points, and ultimately a method of using Bottom-up visual attention model detected region of interest. Please view the full Q +: 351 916 072 Gets
Key words image retrieval; visual attention model; region of interest; Itti model
2.3.1 human visual perception and cognitive processes
This section from cognitive psychology of human vision to discuss the process of perception.
Receptor, effector, memory and processors to human information processing system. Susceptor to receive external information, effector react internal representation in external memory and storing the extracted information present in the form of symbols and structure, creating said processor performs symbols and symbols Xi structure outside flag information, copy, modify and destroy and other operations. After the human visual perception system combines the many theories of cognitive psychology in the formation. The following will separately from visual perception, and memory configuration information selecting these three classes line brief description thereof.
Figure 2.3 Examples of visual attention mechanism
for geometry in Figure 2.3, we can very clearly feel the presence of visual attention mechanisms. Ring hollow, solid black dots, the line will be different directions quickly caught our attention. These are the visual attention mechanism at work. Various visual features are extracted in a stage preceding note parallel out in a serial fashion and integrate visual attention in object stage. Note that the preceding phase extracted visual feature also referred early visual features.
In the choice and transfer the focus of attention, Kouch (1985) conducted in-depth research, he put forward the focus of attention (FOA) change has the following characteristics:
. ① scalability (Zoom lens): spatial extent of the FOA can be enlarged or reduced.
. ② neighboring priority (Proximity): prefer the current look and content accessible position FOA transfer.
. ③ focus shifts (Refocus): FOA can be transferred from one location to another.
. ④ IOR (Inhibition of Return): FOA transfer inhibition when asked recently been selected to return the gaze content.
Selective visual attention is a complex mental activity, which involves a variety of environmental factors, sensation, perception, knowledge and memory. It is broken down into different categories, then it is necessary to study separately. Starting from various angles will get a different classification results.
2.3.2 visual attention physiology model
there are a variety of visual attention models in the field of psychophysics, their purpose is to mimic the behavior of dataIn order to better understand and explain the behavior of human visual perception. These models may be implemented on a computing all these physiological models and calculation will be superimposed on the wanted mold some concepts, such as a characteristic graph of FIG products with the concept, and therefore physiological master visual attention model to improve computational model of visual attention understanding of great help. Now we come to introduce specific model of visual attention psychology influential, they are Treisman comprehensive theoretical characteristics and Wolfe's Guided search model.
Summary of the I
the Abstract II
Chapter 1 Introduction 1
1.1 1 The purpose and significance of the issues
the status quo at home and abroad Study 2 1.2
1.3 dissertation contents and the main work of this paper arrangements 4
Chapter 2 biological visual attention and selection mechanism Model Overview 6
2.1 Introduction 6
2.2 structure 6 human visual system physiology
physiological model vision 8 2.3
2.3.1 human visual perception and cognitive processes 8
2.3.2 physiologically visual attention model 9
2.4 10 Summary
Chapter 3 region of interest based on the detected visual attention model 11
3.1 Introduction 11
3.2 typical visual attention model 11
3.2.1 the Itti model 11
3.2.2 Stentiford model 13 is
3.2.3 the HOME model 14
3.3 based on the periphery of the central contrast saliency detection 14
3.4 Based interested detection scheme visual attention model 21
3.4.1 research ideas 21
3.4.2 The overall architecture and implementation 22
3.5 Summary 22
Chapter 4 Improved detection methods based on evolutionary programming area 23
4.1 Introduction 23
4.2 evolution the basic principle of planning 23
4.3 improved region detection method based on evolutionary programming method 24
4.3.1 improved algorithm Implementation framework 24
4.3.2 global significant computation graph 24,
26 is based on a preselected 4.3.3 candidate region extraction method threshold of
4.4 results analysis 27
4.5 28 Summary
Chapter 5, one kind of model 29 based on the improved method for detecting highlighted regions of interest
5.1 Introduction 29
5.2 display improved model structure 29 projecting
metric significant visual 5.3 30
5.3.1 local wavelet transform significant measure 31
5.3.2 improved evolutionary programming based global significance measure 33
5.3.3 33 fusion normalized
select and transfer the focus of attention 5.4 33
5.4.1 34 FOA selection
5.4.2 transfer factor FOA 36
5.5 38 calculated desired visual
5.6 Interest measure 39
5.7 Summary 39
Chapter VI Summary and Outlook 41
References 42
Acknowledgments 44
Appendix 45
Original link: http: //www.51jrft.com/dzxx/dzkxyjs/552.html