A hardware and software toolkit for lensless computational imaging with Raspberry Pi
The fundamental idea behind lensless imaging is to replace a bulky and costly lens with a mask of equivalent size to the image sensor. However, while the lens gathers a considerable amount of light to capture sharp images, the mask alone is insufficient to produce such results. Therefore, it becomes necessary to employ a reconstruction algorithm to best restore the desired image from the limited amount of light and subtle variations collected through the mask.
The primary advantage of using lensless cameras lies in their compactness. Unlike traditional lensed cameras, which require a certain distance between the lens and the sensor for focusing, lensless cameras can operate without a lens.
Lensless cameras are characterized by their compactness and lightness compared to conventional devices. While traditional lenses have lengths of a few centimeters, the distance between a mask and the sensor in a lensless camera would be on the order of a millimeter. This reduction in size naturally leads to a lighter device, as well as a more affordable manufacturing and purchasing cost.
Moreover, the measurements captured in a lensless camera are inherently blurred and unclear, enhancing privacy levels, as reconstructing the image would require precise knowledge of the camera's parameters.
In lensless imaging, the dimensions of the mask and the sensor are identical, opening the door to the development of sensors with unconventional proportions, such as elongated sensors. This technology offers adaptability and flexibility that conventional imaging does not possess.
A subtle yet significant advantage of lensless imaging lies in the balance between resolution and field of view, a particularly essential consideration in microscopy. Unlike conventional imaging methods where the field of view can be limited, lensless imaging offers a field of view limited only by the size of the sensor. This means that high resolution can be achieved across the entire captured image, a crucial aspect for microscopy applications where precision and visual scope are paramount.
Finally, lensless sensors can be used in compressive sensing approaches because they do not focus light from one point to another but map a point in the scene to multiple points on the sensor. This allows for the reconstruction of higher-resolution images than the sensor's resolution or higher-dimensional quantities, such as 3D imaging or hyperspectral imaging, or the reconstruction of videos from 2D instantaneous measurements.
However, despite the reduction in size, weight, and cost by removing the lens, it comes with a major drawback: image quality. The sensor's size limits light collection, resulting in unclear images during capture and imperfect reconstruction. Additionally, the dependence on computational reconstruction algorithms, which can be time-consuming (in the case of iterative algorithms) or require extensive training data (in the case of machine learning methods), often limits real-time imaging. These reasons make lensed cameras the preferred method for now.