Skip to content

Latest commit

 

History

History
49 lines (27 loc) · 2.73 KB

vevid.md

File metadata and controls

49 lines (27 loc) · 2.73 KB
Raw

Vision Enhancement via Virtual diffraction and coherent Detection (VEViD)

Code Architect

class VEVID:
    def __init__(self, h=None, w=None):

    def load_img(self, img_file=None, img_array=None):

    def init_kernel(self, S, T):
    
    def apply_kernel(self, b, G, color=False, lite=False):    

    def run(self, img_file, S, T, b, G, color=False): 
    
    def run_lite(self, img_file, b, G, color=False):

  • To instantiate a VEVID class, three parameters h (height), w (width). h and w are set to None by default. If you leave them as None by default, their value will be determined when calling the following load_img method.

  • The load_img method can load the image from either an image file indicated by img_file or an image array indicated by img_array. Then it converts the image to from RGB to HSV. If h and w are not indicated in the __init__ method, they will be determined by the shape of the image. Otherwise, the image will be reshaped to the indicated size.

  • The init_kernel method initializes the VEViD kernel according to the parameters phase scale S and phase variance T. The explanation of the parameters can be found in [11].

  • The apply_kernel method applies the initialized kernel to V channel of the HSV image with a regularization constant b and a phase activation gain G. color is a boolean flag to indicate whether we run VEViD for low-light enhancement or color enhancement. lite is a boolean flag to indicate whether the approximated version of VEViD is used. The explanation of the parameter, low-light enhancement vs. color enhancement and the approximation can be found in [11].

  • The run method runs the full VEViD algorithm which wraps load_img, init_kernel, apply_kernel together.

  • The run_lite method runs the accelerated VEViD algorithm called VEViD-lite which wraps load_img and apply_kernel together.

GPU Accelerationn

The GPU version of VEViD significantly accelerates the PST algorithm by using GPU(s). As defined in phycv/vevid_gpu.py , the architect of the VEVID_GPU class is similar to the original VEVID class with same attributes and methods. The main differences are:

  1. You have to indicate the device to operate on when instantiating the class.
  2. The image is loaded as torch.Tensor instead of numpy.ndarray.
  3. The returned result also locates on GPU and is in the form of torch.Tensor.

Examples

Example 1 - Single Image ( scripts/run_vevid.py )

In this example, we run VEViD on a single image and all steps are performed in a single run method.

Example 2 - Video ( scripts/run_vevid_video.py and scripts/run_vevid_lite_video.py )

In these examples, we run VEViD and VEViD-lite on a video.