Verification of a modular graph based image processing system
Lehtinen, Matias (2016)
2016
Sähkötekniikan koulutusohjelma
Tieto- ja sähkötekniikan tiedekunta - Faculty of Computing and Electrical Engineering
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Hyväksymispäivämäärä
2016-05-04
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tty-201604203826
https://urn.fi/URN:NBN:fi:tty-201604203826
Tiivistelmä
Electronic devices today have become complex. Any non-trivial device consists of both hardware and software. Tightening time to market and cost requirements put pressure on the development process of the devices. Software and hardware needs to be developed concurrently and must be verified in an early phase of product development.
This thesis introduces a graph based image processing system. Image processing system is a complex system that usually consists of software, firmware and hardware. The possibilities and methods of graph verification are investigated in this thesis. Graphs can be used to handle the complexity of the system by encapsulating the functionality of the underlying implementations. Graphs provide modularity and configurability that can be utilized in the development and verification of the system. Reuse of software is increased due to the consistent and defined nature of graphs and their vertices. Software development shift left can be enabled by performing graph vertex verification in isolation by using pre-silicon development platforms.
In this thesis, image processing system graphs were also used in a real life product development project. Graph verification was initiated early in the product development. Shift left was exercised by utilizing the graph verification in several pre-silicon platforms. Functional, performance and stability testing was implemented. Both complete graphs and their vertices were verified in isolation. Graph verification provided many benefits to the product development. Implementations could be tested in several different environments in isolation using only a light test framework. Issues could be found and fixed early. Performance bottlenecks could be pinpointed and acted upon.
With the foundations laid in this project, it would be possible in the future to take more advantage of graphs. More advanced automated image quality testing would allow efficient verification. Finer granularity graphs would allow more configurability and more focused testing. Shift left could be further increased by adapting the development of the algorithms to use graphs. This would lower the gap between algorithms and actual vertex implementations and also introduce the available test infrastructure to algorithm development.
This thesis introduces a graph based image processing system. Image processing system is a complex system that usually consists of software, firmware and hardware. The possibilities and methods of graph verification are investigated in this thesis. Graphs can be used to handle the complexity of the system by encapsulating the functionality of the underlying implementations. Graphs provide modularity and configurability that can be utilized in the development and verification of the system. Reuse of software is increased due to the consistent and defined nature of graphs and their vertices. Software development shift left can be enabled by performing graph vertex verification in isolation by using pre-silicon development platforms.
In this thesis, image processing system graphs were also used in a real life product development project. Graph verification was initiated early in the product development. Shift left was exercised by utilizing the graph verification in several pre-silicon platforms. Functional, performance and stability testing was implemented. Both complete graphs and their vertices were verified in isolation. Graph verification provided many benefits to the product development. Implementations could be tested in several different environments in isolation using only a light test framework. Issues could be found and fixed early. Performance bottlenecks could be pinpointed and acted upon.
With the foundations laid in this project, it would be possible in the future to take more advantage of graphs. More advanced automated image quality testing would allow efficient verification. Finer granularity graphs would allow more configurability and more focused testing. Shift left could be further increased by adapting the development of the algorithms to use graphs. This would lower the gap between algorithms and actual vertex implementations and also introduce the available test infrastructure to algorithm development.