Neural Program Synthesis for Automatic Image Enhancement
Program synthesis is the art of synthesizing programs from the user's specifications. The user's in-tent is specified through demonstration, natural language, input-output examples, and so on. Nowadays, Program Synthesis has been applied in image manipulations also. The proposed sys-tem ai...
Gespeichert in:
| Veröffentlicht in: | Journal of internet services and information security Jg. 15; H. 2; S. 791 - 806 |
|---|---|
| Hauptverfasser: | , , , , |
| Format: | Journal Article |
| Sprache: | Englisch |
| Veröffentlicht: |
30.05.2025
|
| ISSN: | 2182-2069, 2182-2077 |
| Online-Zugang: | Volltext |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| Zusammenfassung: | Program synthesis is the art of synthesizing programs from the user's specifications. The user's in-tent is specified through demonstration, natural language, input-output examples, and so on. Nowadays, Program Synthesis has been applied in image manipulations also. The proposed sys-tem aims to synthesize programs on image enhancements by replicating the manipulations done in the given image. Our synthesizer is a hierarchical learning system, which predicts the type of manipulations on images like rotation, saturation, contrast, blur, filter, resize properties of the pic-ture. Additionally, the system predicts the parameter information associated with every function. The Convolution Neural Network has been used for both function prediction and parameter pre-diction. The third layer of the system is an expert system, which generates python programs when an edited input image is given. Though we have automatic image editing tools like Adobe Auto Enhance or Google Photos, AI editor, our Neural Program Synthesis is transparent with the code. The expert system permits the user to update the system on different demands. The system pro-vides overall accuracy of 94 % over different image editing operations. |
|---|---|
| ISSN: | 2182-2069 2182-2077 |
| DOI: | 10.58346/JISIS.2025.I2.052 |