# US 2025/0095210 A1 · Zero-Vision Camera System

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2025/0095210 A1
- Document ID: US20250095210A1
- Status: Published application
- Kind: A1
- Application number: 18/604,683
- Filed: 2024-03-14
- Published: 2025-03-20
- Earliest claimed: 2023-09-16
- Assignees: SiliconeSignal Technologies
- Applicants: SiliconeSignal Technologies
- Inventors: Khalid Saghiri
- Web page: https://patentgrep.com/patent/US20250095210A1

## CPC

- G06T 9/00: Image coding (bandwidth or redundancy reduction for static pictures H04N1/41; coding or decoding of static colour picture signals H04N1/64; methods or arrangements for coding, decoding, compressing or decompressing digital video signals H04N19/00)
- G06T 5/10: using non-spatial domain filtering
- G06T 5/20: using local operators
- G06T 5/50: using two or more images, e.g. averaging or subtraction
- G06T 5/70: Denoising; Smoothing
- G06T 2207/20052: Discrete cosine transform [DCT]
- G06T 2207/20084: Artificial neural networks [ANN]
- G06T 2207/20224: Image subtraction
- G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL

## Abstract

A hardware camera may include a camera sensor configured to determine input image data. The hardware camera may also include an image signal processor configured to perform one or more image signal processing operations on the input image data. The hardware camera may also include a neural processing unit configured to determine encoded image data by encoding the input image data with an image data encoder portion of a camera autoencoder. The camera autoencoder may be trained based on training image data collected from the camera sensor and a fingerprint specific to the hardware camera. The hardware camera may also include a camera communication interface configured to transmit the encoded image data to a remote computing system, which may determine decoded image data by decoding the encoded image data via an image data decoder portion of the camera autoencoder.

## Claims

### Claim 1

A method comprising:
determining a camera autoencoder trained based on training image data collected from a hardware camera and a fingerprint specific to the hardware camera, the camera autoencoder including an image data encoder stored on a storage device at the hardware camera and an image data decoder stored on a remote computing system, wherein determining the camera autoencoder comprises training the camera autoencoder based on the training image data, storing the image data encoder on the storage device, and transmitting the image data decoder to the remote computing system;
determining input image data from a camera sensor included in the hardware camera;
determining encoded image data by encoding the input image data at the hardware camera with the image data encoder, wherein the image data encoder includes a plurality of encoder internal neuron layers and encodes the input image data into a latent space neuron layer; and
transmitting the encoded image data to the remote computing system, the remote computing system determining decoded image data by decoding the encoded image data via the image data decoder, wherein the image data decoder includes a plurality of decoder neuron layers and decodes the encoded image data from the latent space neuron layer into a decoder output neuron layer.

### Claim 2 (depends on 1)

The method recited in claim 1, wherein the fingerprint is a fixed pattern noise (FPN) dark signal non-uniformity (DSNU) pattern.

### Claim 3 (depends on 2)

The method recited in claim 2, the method further comprising determining the FPN DSNU pattern, wherein determining the FPN DSNU pattern comprises:
capturing one or more dark frame images with the camera sensor while the camera sensor is covered.

### Claim 4 (depends on 3)

The method recited in claim 3, wherein determining the FPN DSNU pattern further comprises:
determining a pixel-wise average frame based on the one or more dark image frames; and
determining a denoised average frame based on the pixel-wise average frame.

### Claim 5 (depends on 4)

The method recited in claim 4, wherein determining the FPN DSNU pattern further comprises:
determining an updated frame by subtracting the denoised average frame from the pixel-wise average frame; and
applying a discrete cosine transform high-pass filter to the updated frame.

### Claim 6 (depends on 1)

The method recited in claim 1, wherein the remote computing system includes a deep learning analytics image decoder neural network that includes the image data decoder and one or more analytics neuron layers configured to perform one or more analytics tasks.

### Claim 7 (depends on 6)

The method recited in claim 6, wherein the deep learning analytics image decoder neural network includes a plurality of obfuscation layers collectively configured to implement a transformation and an inverse of the transformation.

### Claim 8 (depends on 6)

The method recited in claim 6, wherein the deep learning analytics image decoder neural network is configured to perform the one or more analytics tasks without producing a visualizable image stream at the remote computing device.

### Claim 9 (depends on 1)

The method recited in claim 1, wherein the image data encoder is entangled in an image signal processing deep learning network implemented at the hardware camera and configured to perform image signal processing operations on the input image data.

### Claim 10 (depends on 9)

The method recited in claim 9, wherein the encoded image data is determined and the image signal processing operations are performed without producing a visualizable image stream at the hardware camera.

### Claim 11

A hardware camera comprising:
a camera sensor configured to determine input image data;
an image signal processor configured to perform one or more image signal processing operations on the input image data;
a neural processing unit configured to determining encoded image data by encoding the input image data with an image data encoder portion of a camera autoencoder, the camera autoencoder being trained based on training image data collected from the camera sensor and a fingerprint specific to the hardware camera, the camera autoencoder including an image data decoder stored on a remote computing system, the image data encoder including a plurality of encoder internal neuron layers and encoding the input image data into a latent space neuron layer; and
a camera communication interface configured to transmit the encoded image data to the remote computing system, the remote computing system determining decoded image data by decoding the encoded image data via the image data decoder, the image data decoder including a plurality of decoder neuron layers and decoding the encoded image data from the latent space neuron layer into a decoder output neuron layer.

### Claim 12 (depends on 11)

The hardware camera recited in claim 11, wherein the fingerprint is a fixed pattern noise (FPN) dark signal non-uniformity (DSNU) pattern, wherein the hardware camera is further configured to determine the FPN DSNU pattern at least in part by capturing one or more dark frame images with the camera sensor while the camera sensor is covered, and wherein determining the FPN DSNU pattern comprises:
determining a pixel-wise average frame based on the one or more dark image frames;
determining a denoised average frame based on the pixel-wise average frame;
determining an updated frame by subtracting the denoised average frame from the pixel-wise average frame; and
applying a discrete cosine transform high-pass filter to the updated frame.

### Claim 13 (depends on 11)

The hardware camera recited in claim 11, wherein the remote computing system includes a deep learning analytics image decoder neural network that includes the image data decoder entangled with one or more analytics neuron layers configured to perform one or more analytics tasks.

### Claim 14 (depends on 13)

The hardware camera recited in claim 13, wherein the deep learning analytics image decoder neural network includes a plurality of obfuscation layers collectively configured to implement a transformation and an inverse of the transformation.

### Claim 15 (depends on 14)

The hardware camera recited in claim 14, wherein the deep learning analytics image decoder neural network is configured to perform the one or more analytics tasks without producing a visualizable image stream at the remote computing system.

### Claim 16 (depends on 12)

The hardware camera recited in claim 12, wherein the image data encoder is entangled in an image signal processing deep learning network implemented at the hardware camera and configured to perform image signal processing operations on the input image data.

### Claim 17

One or more non-transitory computer readable media having instructions stored thereon for performing a method, the method comprising:
determining a camera autoencoder trained based on training image data collected from a hardware camera and a fingerprint specific to the hardware camera, the camera autoencoder including an image data encoder stored on a storage device at the hardware camera and an image data decoder stored on a remote computing system, wherein determining the camera autoencoder comprises training the camera autoencoder based on the training image data, storing the image data encoder on the storage device, and transmitting the image data decoder to the remote computing system;
determining input image data from a camera sensor included in the hardware camera;
determining encoded image data by encoding the input image data at the hardware camera with the image data encoder, wherein the image data encoder includes a plurality of encoder internal neuron layers and encodes the input image data into a latent space neuron layer; and
transmitting the encoded image data to the remote computing system, the remote computing system determining decoded image data by decoding the encoded image data via the image data decoder, wherein the image data decoder includes a plurality of decoder neuron layers and decodes the encoded image data from the latent space neuron layer into a decoder output neuron layer.

### Claim 18 (depends on 17)

The one or more non-transitory computer readable media recited in claim 17, wherein the fingerprint is a fixed pattern noise (FPN) dark signal non-uniformity (DSNU) pattern, wherein the hardware camera is further configured to determine the FPN DSNU pattern at least in part by capturing one or more dark frame images with the camera sensor while the camera sensor is covered, determining a pixel-wise average frame based on the one or more dark image frames, determining a denoised average frame based on the pixel-wise average frame, determining an updated frame by subtracting the denoised average frame from the pixel-wise average frame, and applying a discrete cosine transform high-pass filter to the updated frame.

### Claim 19 (depends on 17)

The one or more non-transitory computer readable media recited in claim 17, wherein the remote computing system includes a deep learning analytics image decoder neural network that includes the image data decoder entangled with one or more analytics neuron layers configured to perform one or more analytics tasks, and wherein the deep learning analytics image decoder neural network is configured to perform the one or more analytics tasks without producing a visualizable image stream at the remote computing system.

### Claim 20 (depends on 18)

The one or more non-transitory computer readable media recited in claim 18, wherein the image data encoder is entangled in an image signal processing deep learning network implemented at the hardware camera and configured to perform image signal processing operations on the input image data.
