When we talk about taking photos, what are we talking about-the evaluation dimension of mobile phone photo quality.

Thanks to the rapid development of computer photography, machine learning and other technologies and their deep application in mobile phone imaging system, the camera performance of smart phones has advanced by leaps and bounds, and the competition among major mobile phone manufacturers in camera quality is like a marathon.

Different from the marathon, there is only one indicator-when running the whole course, when we talk about the quality of a photo taken by mobile phone, what specific dimensional indicators should we mainly look at? That is, how to evaluate the quality of a photo?

Note: This paper only discusses the quality of the final film of the mobile phone, and does not consider the use experience in the process of taking pictures and imaging.

Next, describe several specific aspects of evaluating photo quality:

The exposure of photos includes dynamic range, contrast and exposure repetition accuracy. The quality of exposure control reflects the camera's ability to adjust and shoot according to the brightness of objects and backgrounds.

In some difficult lighting scenes, such as backlight, strong light, dark light and scenes with strong contrast between light and dark, overexposure or underexposure may occur, resulting in the loss of details in bright or dark parts of the picture.

Color is human's visual perception of an object when it is illuminated by light or emits light by itself. In HSV color space, color is represented by hue, saturation and brightness.

The sharpness of a photo indicates the sharpness of each detail shadow and its boundary.

Moderate sharpening will make the picture clearer. Sharpening people's facial features and hair can attract people's attention. However, excessive sharpening destroys the original smoothness, and black and white edges with higher contrast than the real objects appear at the edges of the picture, which affects the quality of the picture.

Some noise reduction techniques (such as longer shutter time and post-processing) may blur the motion of the object, and softening treatment will reduce the surface details of the object, resulting in the inability to accurately restore the surface texture features of the object.

Noise, also known as image noise, is a random change of brightness or color information in an image (the object itself does not). Noise may come from the light of the scene itself, or it may be caused by the electronic components of the image sensor and camera.

Texture and noise affect each other in image processing: image processing that eliminates noise will also reduce details, erase texture and make the image look smoother.

Stray light is unnecessary noise (light), which is caused by imperfect optical machine structure, light source or optical parts outside the field of view, or thermal radiation of the optical or lighting system itself.

Also known as "purple edge", it refers to the phenomenon that color spots appear at the junction of high-light and low-light parts because of the large contrast of the objects photographed by the camera, which is related to lens dispersion and signal processing algorithm.

Due to the inherent characteristics of optical lens (convex lens will condense light, concave lens will diverge light), the object in the photo will be distorted relative to the real thing.

In a photo, there are clear parts and fuzzy parts. Blur includes foreground blur and background blur. The imaging of the blurred part is called defocusing imaging (defocusing). In fact, whether defocus imaging is good or not is the quality of blurred image. Due to the limitation of thickness and space, the camera system of mobile phone can't achieve the shallow depth of field effect in the optical sense of digital camera, so it can only simulate the Jiao Wai blur of traditional camera by blurring the background of photos in the later stage through image algorithm.

The camera needs to focus on the subject accurately under any illumination and motion conditions to make the image of the subject clear. This is a great test of the speed, accuracy and reliability of the autofocus system of mobile phone cameras.

No. Sometimes the photos we want show what we want to see, not what we really see. For example:

No, the purpose of photography is not to show better picture quality, but to better express the photographer's thoughts. So we will deliberately change the image quality to create a certain visual effect we expect, such as:

Better hardware specifications are like better running shoes and sports equipment; Algorithm training again and again is like running training again and again; The score on DxOMark and the user's love are just like our achievements in the marathon and the cheers of the audience.

As Haruki Murakami said in "What do I talk about when I talk about running": "After running a marathon, you will understand that it is not particularly important for a runner to beat or lose to someone in the race. For long-distance runners, whether they can feel proud or something like pride when they finish the race may be the most important. " We strive to improve the quality of mobile phone photos, not only to pursue one point and two points higher than our competitors in the DxOMark ranking, but more importantly, we hope to bring users a better photo experience so that everyone can enjoy the fun of photography!

[1] how does dxomark rate smartphone cameras-analysis-dxomark camera evaluation-dxomark

[2] Jonathan B Phillips & Henrik Eliasson "Camera Image Quality Benchmark"

[3] China Photography Publishing House "New York Institute of Photography Photography Teaching Materials (Volume I)".

[4] IQ 1。 How to define image quality? How to evaluate the image quality? -Zhihu

[5] Huatai Securities "Research Report on Multi-function Superimposed Multi-scene Excellent Long Optical Track-Optical Industry Chain"