• The present invention describes a method & an apparatus of real time restoration of images captured in extreme low-light condition. (Refer Fig.1 & Fig.2)
• Said method comprises a few steps explained and illustrated in figures:
• In First step explains about capturing at least one image in an extreme low-light condition.
• The following step describes about the computing an amplification factor based on one or more parameters.
• Yet further step explains about the amplifying at least one image based on the computed amplification factor.
• Still next step describes about selecting one or more scale levels from an image scale space, wherein each scale level has a respective high, medium, and/or low resolution with respect to one another, wherein said selection is performed by a by a scale selection unit, and selecting the one or more scale levels from image scale space comprises skipping one or more intermediate scale levels of the image scale space.
• Further step describes about the parallelly encoding, by one or more encoders, the at least one amplified image at the one or more selected scale levels to restore the captured image with enhanced quality.
• The encoding the at least one amplified image at higher scale levels comprises simultaneously processing a rectified output and a non-rectified output of convolutional layers present in modified Residual Dense Blocks (RDBs).
• The modified RDB facilitates better restoration & maintains time-memory complexity efficiently.(Refer Fig.3)

APPLICATIONS:

Proposed invention is applicable in the area of below and related sectors:

• Smart phone cameras,
• DSLR cameras,
• ADAS and automobiles
• Image Editing/Enhancement software,
• Photoshop,
• Computer Technology.