The text on the picture is blurry and hard to read, how can I make it clear with one click?

The text content in the image usually contains very important information, however, the text content may be blurred due to shooting limitations, low resolution, the subject is too far away, etc. For example, when taking photos and uploading invoices, due to inaccurate focus, lighting and other problems, the App cannot automatically identify the content of the invoices; when some files are to be uploaded to the background database, some texts are unrecognizable due to their age; they have been saved and forwarded many times. Screenshots of , due to the different compression levels of each application, become blurred when sending to friends or social platforms. In order to improve text resolution and bring users a good user experience, Huawei HMS Core's machine learning service text and image super-resolution capabilities can help.

1. Business Introduction

HMS Core Machine Learning Service Text and Image Super-Resolution Service can magnify images containing text content by 3 times, and at the same time significantly enhance the clarity and recognizability of the text in the image, and easily solve the problem of low text resolution in the image.

DEMO demo

before the score

After overscore

2. Scenario introduction

Text and image super-resolution services are widely used in life. Sometimes when we take a photo, the shooting effect is not ideal due to reasons such as long distance and out-of-focus. This function can be integrated with a photo-taking or image processing app. For some old books or files, the text content is becoming blurred and difficult to identify, and it needs to be uploaded to the database. It can be integrated in its own App, such as library, e-book and other apps, to enhance the text in the picture. Clarity, improve the recognizability of text, and keep knowledge forever.

3. Features

The effect is obvious : 3 times magnification of the picture containing text content, the common jpg compression or downsampling image has obvious super-score effect.

Fast : At present, the algorithm is developed based on deep neural network, making full use of the NPU chip of Huawei mobile phone to accelerate the neural network, and the acceleration ratio can reach more than 10 times.

Portable : Using this API, you can greatly save the time for algorithm development, save the ROM space occupied by the algorithm model, and make your application more portable.

4. Precautions

• Before using the text-image super-divided rate service, you need to convert the image into bitmap in ARGB format, and output the bitmap in ARGB format after processing by this service.

• For common jpg compression or downsampling images, the over-score effect is more obvious, but for pictures with better input image clarity, the effect may not be obvious.

• The text and image super-resolution service requires that the maximum resolution of the input image is 800px*800px, and the minimum size of the input image is not less than 64px on the long side.

Learn more details>>

Visit the official website of Huawei Developer Alliance
Get development guidance documents
Huawei Mobile Services Open Source Warehouse Address: GitHub , Gitee

Follow us to know the latest technical information of HMS Core for the first time~