top of page
  • Writer's pictureMirai vision TAGUCHI

Extract the face and correct the size and brightness[en]

Aligning facial images is a difficult task

In recent years, various web tools and applications have come to be used due to the activation of intra-organizational collaboration represented by Microsoft Teams in companies.

This can annoy its administrator.

It is to collect photographs of the faces of employees.

No, most companies already have photos of their employees ready.

However, in most cases, most of the photos show the upper body of employees.

It is very difficult to align only the face part with the same size from the image of the upper body.

This is because, depending on the time of day, the person taking the photo and the location will be different, so it will not be possible to take exactly the same photo.

Then, it is necessary to enlarge the face part of the person who is a little far away, and similarly, if it is too close, it is necessary to reduce it.

Also, depending on the shooting location, the face may appear dark or too bright.

Furthermore, depending on the web tool or application, the image file size (pixel range specification) is specified differently.

To do this manually requires a great deal of perseverance and skills such as web tools.

ASURA can solve it in one shot!

Please refer to the button below for the operation method through the whole.

The most important part is explained here.

Drag and drop a folder containing image files

Image files of employees, etc. are usually stored together in some folder.

That folder may have many subfolders.

Collecting them one by one is a difficult task.

But don't worry.

In ASURA, if you drag and drop the top folder, all the jpg and png files in the subfolders under it will be referenced.

Also, there may be various other files in that folder, but there is no problem because all files other than jpg and png files will be ignored.

Determine output image size

Decide the output size of the face image file you want to output.

You can output from 80 pixels to 500 pixels.

Decide how close you are to your face

Decide whether you want your face to fill the image file or step back a little and output as if it were photographed.

I will give a concrete example.

"Face proximity = 0.1, 0.2, 0.3, 0.4, 0.5" in order from the left image.

Decide on the brightness of the face

Depending on the location where the face is photographed, some people may have had a dark face.

In such a case, change the setting value here.

A contrast value of 0.0 is used as a concrete example.

"Face brightness = 0.2, 0.4, 0.6, 0.8, 1.0, 1.2, 1.4, 1.6, 1.8" in order from the upper left image.

Decide on the brightness of the face

Brightness can also be adjusted by contrast.

A brightness value of 1.0 is used as a concrete example.

"Face contrast = -80, -60, -40, -20, 0, 20, 40, 60, 80" in order from the upper left image.

Click the "Execute" button

A setting screen that cuts out face images by ASURA.
A setting screen that cuts out face images by ASURA.

Let's wait for the skateboard performance


When the performance is over, the folder where the face image file was output is automatically displayed.


MIRAI VISION OFFICE is making surveillance camera software "ASURA" with face recognition and recording function.

We would like to make applications using AI as daily necessities so that anyone can use them easily.

4 views0 comments


bottom of page