Image synthesis

Overview

includes: color synthesis, synthetic image color consistent with the natural color of the original viter; fake color synthesis, synthetic image color is different from the natural color of the original .

General processing flow

This type of technology is one of the actual application of media computing technology. It is based on the Internet, the image desired element object given by the user. And the specific feature of the object, by a certain screening filter algorithm, select the most matching image source from the image library, and then subjected to segmentation, extraction of the image object, and finally synthesize the result image that can satisfy the user needs after processing.

The general processing process is:

1. Users give a sketch of the expected image, automatically download images from the Internet from the Internet, based on text tags Image filtration of contours and content consistency, dividing the foreground image

3. Calculate the optimal image synthesis combination

4. Synthesate the top ten combination of training, Provide the user to the user, the user can select a satisfactory synthetic result and make a simple interaction correction.

Core Technology

In this process, the core technology used is:

l Image retrieval and image filtering

According to the sketch and text tag given by the user, the system needs to automatically filter and download the appropriate image in the Internet image library, which requires the system to pre-processing the images in the Internet image library, extract the characteristic information of the image, then The information input to the user is compared with the image that is in line with the feature input from the user to make further processing.

The characteristic extraction of the image, introduces a variety of extraction algorithms in this lecture, such as angular points extraction, SIFT feature extraction, regional extraction, etc., the system can choose different extraction methods according to different situations.

The search can also be implemented according to the actual situation to select the appropriate search algorithm.

L Image Segmentation and Object Extract

After retrieving images that conform to the user description characteristics, it is necessary to further process, segment, and extract the object elements required by the user, as the last image. Synthetic infrastructure unit.

image segmentation technology, in this lecture, the main idea has sports clustering and color clustering, and there are also some more mature algorithms.

The extraction of the object, focusing on the concept of the mask in this lecture.

L Image Synthesis

After extracting an object element that conforms to the user description character, the last step is to combine each independent element into an image and then feedback to the user.

The most critical technology in this process is image synthesis technology. In the lecture, image synthesis technology based on hybrid convenient conditions is introduced, compared to the previous method, less synthetic, while also makes each The illumination of the object element is consistent, so that the image is really stronger.

Based on Internet image synthesis technology, it is an emerging technology application that emerged in the large environment of the Internet. The biggest advantage is that it makes full use of the Internet massive image data, great rich As the image object element library as a synthetic candidate source, the significant user needs can be satisfied. At the same time, as the Internet image data is increasing, the object elements will continue to enrich, which allows the system to achieve user needs, and support more complex and more personalized users.

With the rapid growth of the data, the system's requirements for processing technology will be higher and higher, and some faster and more accurate processing technologies will continue to be applied to the system, which will also As a source of power, the continuous development of the media calculation technology is constantly promoted.

Image Synthesis Classification

Modeling

1, one element synthesizes in the picture, size design, location placement;

two The effect of an element synthesizes to other elements in the picture. Main use tools: Dismoring tool set.

Brightness Match

The object of the same texture will present the same diffuse reflection in the same light-receiving environment.

Main use tools: curve, color step, brightness contrast, etc. Adjust the image first adjust the brightness, then color. Because the brightness has only one value, the color has three values. The image first adjusts the saturation to zero, re-adjust its brightness.

Color match

Color difference is mainly caused by different or different post adjustments of the light-receiving environment. The former differences can be achieved by curves, color steps, color balance, color phase saturation, photo filter or even mixing mode, while the latter difference requires a very strong observation and capacity.

Light matches

Different bright and shadows due to the different prospects and backgrounds of light. These are the most difficult to modify in the later period. We cannot change the light-receiving position of the object, and can only rotate the oscillated object to simulate consistent light. Due to the same properties of the object, it can also be divided into opaque, transparent, transparent, opaque (such as grape, paper, etc.), opaque objects diffuse reflections and mirror reflections, transparent objects have refractions and focal characteristics, more There is also a environmental effect between the objects, which require designers to observe.

Light matching should pay attention to 6 points: diffuse reflection, mirror reflection (note water surface reflection), fuzzy reflection, refractive, luminescent, and ambient color.

element quality matching

The surface of different objects has different performance characteristics, pay attention to the fusion of the texture of the texture when synthesizing. Never use the brush and the mask to brush a matter, you can only deceive yourself, you must do the normal fusion of texture (also use masks and brush, but you have to rub along the texture).

screen texture Match

The material taken by different equipment has different textures (such as a unique particle sense of film), the same equipment is also equipped with different light-receiving environments and ISO The texture, exactly the same shooting environment and camera parameters may also be different texture by the backward adjustment of the rings, even if the same picture is in the bright and dark portions. The general texture matching principle is: high quality matching low quality.

The depth of field matches

The depth of field has a spatial sense and a protruding body. The depth of field has three main elements: aperture, lens focal length, and shooting distance. The size of the film will also have a certain impact on the depth of field. So when synthesizing, you need to pay attention to these factors to create the correct synthesis. At the same time, the fuzzy morphology is not lens, and the non-Gaussian blur can be simulated, and the blurred effect is required to match the fuzzy form or even control the focus position. The focus area of ​​the camera is a focus plane, the non-focus, the depth of field is mainly the front and rear relationship, and there are very few left and right relationships.

Distortion Match

The same focal length will generate a distortion of the same degree. The distortion performance in the wide-angle lens will be more obvious. Despite the distortion of the picture edges, it is harmful to the distortion. Distortion produces excessive perspective, often uses more content and creates stronger space sensation, the fisheye lens is the special effect of using distortion. When dealing with distortion, you need to pay attention to the edge distortion and the center is the same, and the edge will be strong. The picture observed by the AE camera is only in vision without distortion. This needs attention when synthesis.

Dynamic Match

In the case where the shutter speed is relatively low, the object sports will produce motion blur. Sports blurred size irreident motion speed and shutter speed. Under normal circumstances, the photographic apparatus can be blurred by the limit of the number of frames of the frame by accelerating the shutter speed solidifier object. On the other hand, if the camera moves, then simultaneously generates motion blur, it is necessary to track the synthetic object to track the operation, then match the camera movement. Generally integrated 2D tracking in general synthesis software, there are also some 2.5D, 3D tracking, and can be more complex motion matching operations.

Environmental fog

The fog is caused by a clean air. Theoretically any scene has a foggy existence. The effect of water mist (白) and dust (biased ash) is the same. Great purpose: close to the virtual (virtual non-transparent), near saturation, supersaturation.

Up Reality

Surreal is above the reality. It is a kind of unlikely based on reasonable basis, like a true art, starting from the Fourth International Training Photoshop, the lecture has been in the lecture. More and more feel that the difference in people is in thinking. Some people can become a master a year, and some people have been in a few years, and the depth of thinking decides to get the amount of information. Differences in thinking. If you feel confused and bottleneck when you study, then ask yourself. When you learn a software, there is no thinking of the following six levels:

1, software level. Learn the software, know what the order is meaning. Will achieve some simple production, but the effect is very poor, you have to lead. This is a person who reciting the order of the order;

2, the application level. Know what software can do, very proficient in software function, very well-understood, read the world's effect, do the best in the world. But leaving the reference and steps, but it is still awkward. This is a person who study cases and tutorials of all year round;

3, equipment level. For PS, some mixed mode simulation is film dark house; color simulation is exposure curve and white balance; the filter special simulation is a variety of solid filter. It can be said that each core function corresponds to a real-existing hardware entity. If you understand its corresponding hardware features, then do things are accurate and fast. At the same time, the ability is only a single software, and any similar software can quickly get started. This is a person who loves traditional industries and computer art design, and thinks.

4, the program level. I need to know some algorithms for the software again. Although the equipment and software are connected, the software is analogue to simulate hardware implementation through an algorithm. It is necessary to understand how the mixed mode is calculated, how is it affects the exposure curve, and there is no color or the like. In this way, you can understand that all kinds of toning, which are useful for fools, any problem, only which toning tool can be used, use another, why is it, each characteristic and defect Where is it? Sometimes only the features of the software are 丌 丌, but it is best to analyze the characteristics, so that you will abandon the tutorial, you will have your own tricks and ideas.

5, observation level. I hope to do something is beautiful, but it is beautifully of the work of everyone. Before doing something, do you want to do it, what style, what color, outstanding, weaken, who is dark, who is a little, who is the body is who is accompanying, how to arrange. When the mind is completely conceived, it is achieved, there are several levels of ability to support, just like plagiarism, the next pen is like God.

6, thinking level. What to do is beautiful, and it is just a unforgettable person, and customer needs is personal guidance. The ultimate work is the most precipitial technology, but incorporates the artistic masters to social thinking about life. The designer is doing someone else, the master is doing her own things, because the level of thinking is the same. Art is not philosophical, and the depth of personal thinking determines the depth of work and life.

Follow the wrong

Problem: Refraction

Match Found (5 photos)

Question: Shadow

Question: Shadow

Problem: Movement

Problem: Light

Image Synthesis Technology

Background Fuzzy Law

This method is to select the background portion and directly blurred. It is very simple to operate.

Spotlights

The main enlargement method

(the main body is properly enlarged, you can zoom in to the local ring. There is a feeling of the cranes).

Color Guidance Method

(When introducing the product of the personnel, the color is more messy can appropriately change the other part to the color identity color of the main body near the main body. Wait).

Contour line graphical illustrated

(this kind of discipline is characterized by using the coherent part of the line, while the body uses the active prototype, the effect is not general) .

Local dilution method

(this 斱 method is protocol to some of the full object, and others are delayed).

Cutting color method

(in order to highlight a part of the physical ringman, this part is cut down, and then appropriately beautify the process).

Line Guidance method

(in the need to protrude with the line ring arrow arrow, etc., quickly attract readers' attention).