Visual perception feature 3 image significance analysis based on visual attention

Tomorrow will be two, and this afternoon also wayward play, play the ball also to do 30 push-ups, now feel arm a bit sour. Wish me luck tomorrow. Today is the last day of this month, come on.

According to the visual perception of the "1", visual attention is an important feature in the information processing mechanism of human visual system, so it is important to study the image significance analysis based on visual attention.

The human vision system is inherently capable of dealing with complex scenes, which quickly scans the entire scene and directs the eyes to the most relevant areas, accurately filtering out redundant, secondary visual information and then quickly locking down important targets. It mainly has two modes, one is the bottom-up model, it is an unconscious, spontaneous response to the scene stimulation, it depends only on the content characteristics of the scene, not with the viewer's memory, consciousness, the main factors are the underlying image features, the other is the top-down model, subject to consciousness, dependent on specific commands, The main influencing factors are high-level semantics, including some prior knowledge in brain memory, fixed module matching and so on.

(1) The underlying image features include:

Contrast: Human vision systems can quickly lock out areas of image brightness contrast in the early stages of visual attention. Those pixel areas that are different from the surrounding area will attract more visual attention. For example, the red flowers in a green meadow will naturally arouse people's visual attention.

Shapes: People tend to be more interested in the shape of something that is more specific than
An object of almost the shape around it. In addition, the pixel points on the target edge are also easy to attract people's attention.
Other low-level image features that affect visual attention include size, color, motion, position, and so on.

(2) High-level semantic factors include:

Contextual content: When viewing an image, the brain controls visual attention, making the human eye focus on the contextual content that reflects the current scene, and finding areas of interest in the image to give feedback to a deeper layer of the visual cortex for further processing.
Location: Eye movement experiments indicate that the human eye is more inclined to focus on the center 25% area of the image and look for targets in the region. In other words, the image center area is more likely to attract people's visual attention.
Foreground/background: Visual attention experiments indicate that human eye vision often looks for targets in the foreground area and ignores the contents of the background area.
People: The visual system is very concerned about the human information, such as human face, body and other information will quickly attract people's attention. The areas where people's own information is present are often highly significant.

Eh, but many factors are difficult to quantify, here can only choose brightness contrast, target edge and image center of the three factors.

All right, just write it down here ~ ~ ~