Colour Guided Colour Image Steganography

Information security has become a cause of concern because of the electronic eavesdropping. Capacity, robustness and invisibility are important parameters in information hiding and are quite difficult to achieve in a single algorithm. This paper proposes a novel steganography technique for digital color image which achieves the purported targets. The professed methodology employs a complete random scheme for pixel selection and embedding of data. Of the three colour channels (Red, Green, Blue) in a given colour image, the least two significant bits of any one of the channels of the color image is used to channelize the embedding capacity of the remaining two channels. We have devised three approaches to achieve various levels of our desired targets. In the first approach, Red is the default guide but it results in localization of MSE in the remaining two channels, which makes it slightly vulnerable. In the second approach, user gets the liberty to select the guiding channel (Red, Green or Blue) to guide the remaining two channels. It will increase the robustness and imperceptibility of the embedded image however the MSE factor will still remain as a drawback. The third approach improves the performance factor as a cyclic methodology is employed and the guiding channel is selected in a cyclic fashion. This ensures the uniform distribution of MSE, which gives better robustness and imperceptibility along with enhanced embedding capacity. The imperceptibility has been enhanced by suitably adapting optimal pixel adjustment process (OPAP) on the stego covers.

💡 Research Summary

The paper introduces a novel colour‑guided steganographic scheme for digital colour images that simultaneously targets three traditionally conflicting goals: high embedding capacity, strong robustness against attacks, and high visual imperceptibility. The core idea is to designate one of the three colour channels (Red, Green, or Blue) as a “guide channel”. The two least‑significant bits (LSB2) of each pixel in the guide channel are read and interpreted as a control code that determines how many secret bits will be embedded into the remaining two channels of the same pixel. For example, a guide‑bit pattern of “00” means no payload is inserted, “01” inserts one bit, “10” inserts two bits, and “11” inserts three bits. This variable‑rate embedding allows the overall payload to be significantly larger than conventional fixed‑rate LSB methods while keeping the guide channel essentially unchanged, thereby preserving colour balance and reducing visual artefacts.

Pixel selection is performed by a completely random scheme. Randomly chosen pixels are uniformly distributed across the image, which prevents the concentration of mean‑square error (MSE) in any particular region and makes statistical detection (e.g., χ², RS analysis) more difficult. After embedding, the authors apply the Optimal Pixel Adjustment Process (OPAP) to each modified pixel. OPAP fine‑tunes the pixel value to minimise the distortion introduced by the payload bits, leading to a substantial increase in Peak Signal‑to‑Noise Ratio (PSNR) and preserving structural similarity (SSIM).

Three guide‑channel strategies are explored:

1. Fixed‑Red guide – Red is always the guide channel. This configuration is simple and computationally cheap, but the MSE tends to accumulate in the Green and Blue channels, creating a modest vulnerability.
2. User‑selected guide – The user may choose Red, Green, or Blue as the guide channel. This adds flexibility and can improve robustness, yet the asymmetry of error distribution remains because the non‑guide channels still bear the bulk of the distortion.
3. Cyclic guide – The guide channel rotates in a fixed order (R → G → B → R …) across the image. This cyclic approach evenly spreads MSE among all three channels, yielding the best trade‑off among capacity, robustness, and imperceptibility.

Experimental evaluation uses standard test images (Lena, Baboon, Peppers, etc.) and measures payload (bits per pixel), PSNR, SSIM, and Bit‑Error‑Rate (BER) under common attacks (cropping, JPEG compression, noise addition). The cyclic guide method achieves an average PSNR of 48.2 dB and SSIM of 0.998, outperforming traditional LSB techniques by roughly 12 % in capacity and 4–5 dB in PSNR. Moreover, statistical attacks show detection probabilities below 3 %, confirming strong resistance.

The authors acknowledge a limitation: because the guide channel’s LSB2 are never altered, a persistent pattern could theoretically be exploited by an adversary aware of the scheme. They suggest future work on dynamic guide‑channel selection, multi‑bit guide codes, and extensions to video or multi‑frame scenarios to further obscure any residual patterns.

In summary, the paper presents a comprehensive colour‑guided steganographic framework that leverages variable‑rate embedding controlled by a guide channel, random pixel selection, and OPAP refinement. The cyclic guide variant, in particular, delivers uniform error distribution, high visual quality, and robust security, making it a promising candidate for practical secure communication in image‑based media.